Repository navigation
Releases
v1.9.1
What's New
| Change | What it does |
|---|---|
| Who changed it last | get_controls and check_model_access name the person who last changed each spend limit and Model Access rule, with the time. manage_spend_limit and manage_model_access show it in their before and after views. |
| Error items for apps | When Wiro refuses a run_model run or a send_agent_message message with code 92 (Model Access) or 93 (spend limit), the tool error also carries Wiro's own error items in _meta["ai.wiro/errors"], each with its code and Wiro's sentence, so an app can tell the refusal by its code. The text the assistant reads is unchanged. |
| Links that open the right workspace | The Spend Limits and Model Access links that get_controls, check_model_access, manage_spend_limit and manage_model_access return now name their workspace, so the dashboard opens the limits or rules of the team or personal account they are about, whichever workspace the browser is on. |
Enhancements
manage_model_access: warns when an agent's rule newly stops its built-in voice, so voice notes are no longer transcribed or voice replies go out as text, and says what brings each one back. For an agent that chats on Gemini, the warning follows how its voice notes are transcribed.manage_spend_limit: the note for an agent limit says the agent pauses at its next step once its own limit is used up, even in the middle of a reply. A voice call it is on ends after the current turn, and its scheduled tasks pause until the limit resets or is raised.send_agent_message(Studio and the Workflow Agent): their chat model follows the caller's Model Access rules. When Wiro refuses one with code92, the note lists the chat models that are still open and says whether leavingmodelout helps.get_agent_messageandget_agent_conversation: say when a spend limit or the balance ended a voice call.- Playbooks: the
controls,studio,workflow-agentandagentstopics cover Studio's chat model, an agent's voice, realtime sessions and voice calls that end at a limit, scheduled tasks that pause instead of failing, renewals counted on the wallet and the member who started the plan, and how to find who changed a limit or rule last.
Tools (54)
- Account:
get_account,get_playbook,search_docs - Models:
search_models,get_model_schema,recommend_model,explore,run_model,wait_for_task,get_task,list_tasks,get_task_price,cancel_task,kill_task,upload_file - Workflows:
list_workflows,get_workflow,create_workflow,update_workflow,clone_workflow,delete_workflow,run_workflow,wait_for_workflow_run,resume_workflow_run,list_workflow_runs - Agents:
send_agent_message,get_agent_message,get_agent_conversation,cancel_agent_message,delete_agent_session,list_studio_results,save_workflow_from_agent,list_agent_templates,get_agent_template,list_agent_skills,deploy_agent,list_my_agents,get_my_agent,control_agent,update_agent,manage_agent_billing,get_agent_activity,delete_agent,apply_agent_skills,set_agent_credentials,connect_agent_oauth,disconnect_agent_oauth,manage_agent_custom_skill,list_projects,bind_agent_project - Controls:
get_controls,check_model_access,manage_spend_limit,manage_model_access
Full Changelog: v1.9.0...v1.9.1
v1.9.0
What's New
| Tool | What it does |
|---|---|
get_controls |
Shows the Spend Limits and Model Access rules of the connected workspace: every limit with what it used, what work still running holds and what is left, every rule in plain words, and the spending this month or today by way in, team member, API key and agent. It says whether this connection can change them. |
check_model_access |
Checks one model for one run before anything runs: a run of this connection, a team member's run, a run with an API key, a way in or an agent's run. It names every rule that blocks the model and why. |
manage_spend_limit |
Sets, raises or removes a monthly or daily spend limit for the whole account, a team member, an API key, a way in or an agent, or changes only its alert email. |
manage_model_access |
Updates, replaces or removes a Model Access rule: categories, single models and what every other model gets. |
Every change first returns a summary: what it covers, the value before and after, what was spent this period and when it resets. Nothing is saved until the call is repeated with confirm: true, and a confirm saves only the change its summary showed. If the limit or rule changed in the meantime, a new summary comes back instead.
Enhancements
- Who can change controls: a personal account's owner, and in a team a team admin or the organization's owner. A connection acts as the person who created its project. A team member without admin rights can read the controls but not change them.
controlstoolset and playbook:WIRO_TOOLSETStakescontrols, andget_playbookhas acontrolstopic with recipes: cap the whole account, cap one API key, give each team member a monthly cap, block a category for one member, let an agent use one model, block the LLM gateway, and find why work was refused.run_model,run_workflow,resume_workflow_run,send_agent_messageandupdate_agent: the spend limit note (code93) now says whether the limit is used up, the run costs more than what is left, or work in progress holds part of it. With the controls toolset loaded, the notes for codes92and93point toget_controls,check_model_accessand the change tools.get_accountsays whether this connection can change the controls.- README: the Controls tools, held amounts and the confirm flow.
Tools (54)
- Account:
get_account,get_playbook,search_docs - Models:
search_models,get_model_schema,recommend_model,explore,run_model,wait_for_task,get_task,list_tasks,get_task_price,cancel_task,kill_task,upload_file - Workflows:
list_workflows,get_workflow,create_workflow,update_workflow,clone_workflow,delete_workflow,run_workflow,wait_for_workflow_run,resume_workflow_run,list_workflow_runs - Agents:
send_agent_message,get_agent_message,get_agent_conversation,cancel_agent_message,delete_agent_session,list_studio_results,save_workflow_from_agent,list_agent_templates,get_agent_template,list_agent_skills,deploy_agent,list_my_agents,get_my_agent,control_agent,update_agent,manage_agent_billing,get_agent_activity,delete_agent,apply_agent_skills,set_agent_credentials,connect_agent_oauth,disconnect_agent_oauth,manage_agent_custom_skill,list_projects,bind_agent_project - Controls:
get_controls,check_model_access,manage_spend_limit,manage_model_access
Full Changelog: v1.8.0...v1.9.0
v1.8.0
What's New
| Feature | What it does |
|---|---|
| Model Access | Wiro accounts and teams choose which models can run: for the whole wallet, a team member, an API key, a way in (the dashboard, the API, MCP, the LLM gateway, the browser extensions, agents) or one agent. A rule allows or blocks whole categories (image, video, audio and speech, music, 3D, LLM and chat, utilities) and single models. A run that a rule blocks is refused before it starts with error code 92. |
Enhancements
run_model,run_workflow,resume_workflow_run,send_agent_messageandupdate_agent: a Model Access refusal (code92) comes back as a tool error with Wiro's message, which names the rule and the model, and a note that nothing ran and nothing was charged, not to try again or work around the rule, to tell the user which rule blocked the model, and that whoever manages the rules can change it on the Model Access page.get_accountand the models and billing playbooks name code92. The billing playbook no longer quotes the old team refusals that spend limits and Model Access replaced.
Full Changelog: v1.7.0...v1.8.0
v1.7.0
What's New
| Feature | What it does |
|---|---|
| Spend limits | Wiro accounts and teams can cap what they spend per month or per day: the whole wallet, a team member, an API key, a way in (the dashboard, the API, MCP, the LLM gateway, the browser extensions, agents) or one agent. Work a used-up limit covers is refused before it runs with error code 93. |
Enhancements
run_model,run_workflow,resume_workflow_runand agent calls: a spend-limit refusal (code93) comes back as a tool error with Wiro's message, which names the limit and when it resets, and a note that nothing ran and nothing was charged, not to try again until the limit resets or someone who manages it raises it, and where limits are managed.get_account: lists the workspace's spend limits with what each has used this period and whether it is used up (spendLimitsin the structured result). The balance still answers when the limits cannot be read (spendLimitsError).
Full Changelog: v1.6.2...v1.7.0
v1.6.2
Enhancements
- Paid runs wait for a yes:
run_model,run_workflow,resume_workflow_runand Studio or Workflow Agent turns now tell the user the expected cost in one line and wait for their yes before each run, even when the user asked for exactly that work. A batch still takes one yes for the list or total shown. The server instructions, the playbooks, the tool descriptions and thecreate_with_studioprompt all say so.
Full Changelog: v1.6.1...v1.6.2
v1.6.1
Enhancements
- get_model_schema: hidden parameters are no longer listed. A model that moved a file field to a single combined input keeps its former URL fields working for existing clients, but the schema now shows only the fields to use.
- list_agent_templates: the README lists the three monthly agent plans.
Full Changelog: v1.6.0...v1.6.1
v1.6.0
What's New
| Change | Details |
|---|---|
| Three fixed agent plans | Agents run on Starter, Pro and Max. manage_agent_billing quotes a plan move before it charges, a smaller plan waits for the renewal, and an undo names the plan the agent renews on. |
| Free skill changes | Adding or removing a Skill or Connector moves no money. get_agent_message and send_agent_message show add and remove proposals with what else they change. |
| Add and connect in one step | A proposed Connector comes with its connect card, and a key sent together with the yes is saved right after the add applies, so the agent restarts once. |
| Choose the accounts after a sign-in, in the chat | After a browser sign-in, connect_agent_oauth (status) lists the accounts, Pages, properties or channels it found, and select_accounts saves the ones the user picks. Wiro keeps that list for 15 minutes, only for the person who started the sign-in or token check, and saves only accounts from it. |
| Accounts and their extra step in one save | For Meta Ads (Pages) and Google Business Profile (locations), select_accounts first lists the extra step for the chosen accounts, then saves the accounts and the selections together, so a running agent restarts once. |
Enhancements
- set_agent_credentials: Wiro now shows saved secrets masked, and a write that sends those exact dots is refused, so the mask is never stored as a value.
- send_agent_message and get_agent_message: Wiro's answers in an API chat carry no Wiro page or link, so an integrator's own product can show them as they are.
- manage_agent_billing: a failed card payment points to Billing, and a wallet renewal says what it paid and which plan it took.
- connect_agent_oauth: a Facebook sign-in with the user's own app is ready for the Page choice as soon as the user approves it.
- connect_agent_oauth: a sign-in or token check that did not finish ends the wait and names its error code.
- connect_agent_oauth: the redirect URI to register in the user's own app comes from Wiro, for the host this server talks to.
- connect_agent_oauth, send_agent_message and the agents playbook: the person who deployed an agent or a team admin of its team can connect its accounts. The account list after a sign-in goes only to the person who started it.
- select_accounts: saves accounts by their id from Wiro's list, under the names in that list. Choosing other accounts later takes a new sign-in or token check.
- get_agent_message and send_agent_message: a proposed Connector says how each service it brings connects, also with the user's own app keys or a token, and that a reply does not connect its connect card.
Tools
Models: search_models, get_model_schema, recommend_model, explore, run_model, wait_for_task, get_task, list_tasks, get_task_price, cancel_task, kill_task, upload_file
Workflows: list_workflows, get_workflow, create_workflow, update_workflow, clone_workflow, delete_workflow, run_workflow, wait_for_workflow_run, resume_workflow_run, list_workflow_runs, save_workflow_from_agent
Agents: send_agent_message, get_agent_message, get_agent_conversation, cancel_agent_message, delete_agent_session, list_studio_results, list_agent_templates, get_agent_template, list_agent_skills, deploy_agent, list_my_agents, get_my_agent, control_agent, update_agent, apply_agent_skills, set_agent_credentials, connect_agent_oauth, disconnect_agent_oauth, list_projects, bind_agent_project, manage_agent_custom_skill, manage_agent_billing, get_agent_activity, delete_agent
Account and guides: get_account, get_playbook, search_docs
Full Changelog: v1.5.8...v1.6.0
v1.5.8
What's New
| Change | Details |
|---|---|
| Agents propose setup changes in chat | Ask a deployed agent for a change, such as connecting a connector, adding or changing a scheduled task, saving a standing rule as a custom skill or changing its time zone, and it proposes the change as a setup card. send_agent_message, get_agent_message and get_agent_conversation show each proposal in full: the connector, the task's schedule and instructions, the skill's text or the time zone. Nothing changes until you approve it. |
chat_setup on deploy_agent and update_agent |
Choose where people can approve setup changes from chat: all (the default: the Wiro web chat and messages sent with your API key), webchat (only the Wiro web chat) or off. get_my_agent shows the current value. |
Enhancements
- send_agent_message: a reply that approves or turns down a setup card is answered by Wiro itself, so it starts no agent turn and uses no credits. The tool shows which proposals were applied, dismissed or failed.
- send_agent_message: in a chat sent with your API key, the first "yes" shows the exact change and a second "yes" applies it. The Next step asks you before sending either.
- set_agent_credentials: the key for a proposed connector goes through this tool, never through a chat message.
Tools
Models: search_models, get_model_schema, recommend_model, explore, run_model, wait_for_task, get_task, list_tasks, get_task_price, cancel_task, kill_task, upload_file
Workflows: list_workflows, get_workflow, create_workflow, update_workflow, clone_workflow, delete_workflow, run_workflow, wait_for_workflow_run, resume_workflow_run, list_workflow_runs, save_workflow_from_agent
Agents: send_agent_message, get_agent_message, get_agent_conversation, cancel_agent_message, delete_agent_session, list_studio_results, list_agent_templates, get_agent_template, list_agent_skills, deploy_agent, list_my_agents, get_my_agent, control_agent, update_agent, apply_agent_skills, set_agent_credentials, connect_agent_oauth, disconnect_agent_oauth, list_projects, bind_agent_project, manage_agent_custom_skill, manage_agent_billing, get_agent_activity, delete_agent
Account and guides: get_account, get_playbook, search_docs
Full Changelog: v1.5.7...v1.5.8
v1.5.7
v1.5.6
What's New
| Change | Details |
|---|---|
| A repeated call continues its task on any server | run_model sends a key with each run request, a hash of the caller, the model and the parameters, and Wiro remembers the run under it. When a client drops a long call and the agent repeats it, the repeat continues the same task on whichever server process it reaches, a restarted one included, so one call stays one task. |
Enhancements
- run_model: calls sent side by side still each start their own task, and the same call made again after its answer arrived starts a new one, as before.
- run_model: a continued task that failed is replaced by a new task at most once per call.
- WiroClient:
runModeltakes arunKeyoption, and the newreleaseRuntells Wiro how the last call serving a keyed run ended.
Tools
Models: search_models, get_model_schema, recommend_model, explore, run_model, wait_for_task, get_task, list_tasks, get_task_price, cancel_task, kill_task, upload_file
Workflows: list_workflows, get_workflow, create_workflow, update_workflow, clone_workflow, delete_workflow, run_workflow, wait_for_workflow_run, resume_workflow_run, list_workflow_runs, save_workflow_from_agent
Agents: send_agent_message, get_agent_message, get_agent_conversation, cancel_agent_message, delete_agent_session, list_studio_results, list_agent_templates, get_agent_template, list_agent_skills, deploy_agent, list_my_agents, get_my_agent, control_agent, update_agent, apply_agent_skills, set_agent_credentials, connect_agent_oauth, disconnect_agent_oauth, list_projects, bind_agent_project, manage_agent_custom_skill, manage_agent_billing, get_agent_activity, delete_agent
Account and guides: get_account, get_playbook, search_docs
Full Changelog: v1.5.5...v1.5.6